On Writing
Putting thoughts in public makes my stomach flutter, so this is my attempt to internalize my decision to write more. I’ll pushing myself because when thinking becomes words, they turn around and hold myself accountable. As I write more, I hope to become comfortable enough that I start looking forward to refusals of my arguments, or even turning my thinking upside down.
An immediate benefit of writing I found is the process of structuring and restructuring my thinking. My goal is neither succinctness nor sharing facts. Rather, I hope to washes off clutters in my mind. In doing so, it would be wonderful to also curate a dialogue between whoever is reading and ideas I found interesting.
Among other things, I’ve always wanted to write with a greater sense of responsibility. One of my favorite journalists, John Hersey, once noted that:
If burdens can translate to the weights of stories, a good example is Hersey’s epochal editorial of Hiroshima, arguably one of the most influential pieces of American writing in the 20th century. After the first atomic bomb was dropped over Hiroshima in August 1945, Hersey set out to uncover the effect of nuclear war not to buildings but to humans. He interviewed several dozen survivors and settled on six regular people who shared powerful memories. His sentences put faces into names. And he used narrative to make those faces alive. Doing so, he unmistakenly bled his journalistic core with a kernel of moral preachment. Hersey’s art of facts did wonder — Albert Einstein, for example, requested one thousand copies of the editorial to distribute to members of his Emergency Committee of Atomic Scientists and noted:
Writing with burdens also means writing with a disregard for moral fashion. With the advent of web publishing tools in the 1990s, public writing has been democratized in a dramatic fashion. Everyone write. Just look at twitter. Yet it isn’t hard to notice what people write about is narrowing. Trumps, Kardashians, and Epsteins of the world dominate our screens. Op-ed sections on major newspapers are painfully biased. Different forms of Russel Conjugation in the media exploit the gap between the emotional content of a word and the factual content. Paul Graham wrote about a related thought experiment and called it the Conformist Test:
He made a good point, but this test is not entirely representative. Today, we are often expected to speak. Silence is not an option. Or worse, it translates to a statement. When I last went to a family get-together, I was asked how the food was. “They were all delicious,” I responded without thinking. The dishes were great, but that didn’t matter. In cultures like mine, family unity trumps the dishes. This may be trivial, but the point is societal preferences are preconditioned. And in larger spheres of politics and society, this becomes much more dangerous in aggregate.
In 2016, any party in Silicon Valley would have a discussion about how ridiculous a Trump presidency would be. Whether the party has four people, forty, or four hundred, almost everyone would say they are not voting for Trump. Every poll in the country agrees. Why? Because in Silicon Valley and places alike, there are consequences for saying otherwise so the minority changes their preferences to match the greater consensus. This idea is not new. Nor is the claim that the collective misrepresentation can generate outcomes that are very different from the said public opinion. Economist Timur Kuran calls it “Preference Falsification”. In his 1995 book, Private Truths Public Lies, he puts:
Kuran believes widespread preference falsification is particularly dangerous:
An updated Conformist test would be: Do you have any opinions that you would express the opposite of what you actually think in front of a group of your peers?
Recognizing this problem is an important start, but I wonder if the problem itself will get worse. Last month, OpenAI released the API for its third generation of language prediction model (GPT-3) into the open-source wild. From translation, question-answering, to tasks that require on-the-fly reasoning, such as unscrambling words or using a novel word in a sentence, GTP-3 demonstrated AI can solve many tasks that it has never encountered with just enough data. Just on the text-generation front, humans already can’t distinguish between articles written by humans versus articles written by GPT-3.
The danger, however, is that internet-trained models have internet-scaled biases. When AI starts to generate texts, write screenplays, and produce music, we are in real trouble. Skewed, or worse, falsified input will spit out substances that will become an integral part of our internet fabric.
What now? We need to separate the imperative to write about ideas from ideology and politics. Growing up in China and getting my education at a US university like Stanford, I was often asked about their differences. There’s a lot, but to me their similarity is more striking: the same desire for technological progress and the same resistance to change. Writing what you believe isn’t easy, but we need to start somewhere. Novelist Roberto Bolaño once described why writing is simply necessary.
So this is my attempt to do just that. With that, I just have one ask: Let’s please make this a conversation. If you disagree, let me know. I really do hope to make it better. And if by some miracle I have clearly communicated the point and we actually disagree, maybe we can stop this writing business and move this over to a 1 on 1 conversation.
Putting thoughts in public makes my stomach flutter, so this is my attempt to internalize my decision to write more. I’ll pushing myself because when thinking becomes words, they turn around and hold myself accountable. As I write more, I hope to become comfortable enough that I start looking forward to refusals of my arguments, or even turning my thinking upside down.
An immediate benefit of writing I found is the process of structuring and restructuring my thinking. My goal is neither succinctness nor sharing facts. Rather, I hope to washes off clutters in my mind. In doing so, it would be wonderful to also curate a dialogue between whoever is reading and ideas I found interesting.
Among other things, I’ve always wanted to write with a greater sense of responsibility. One of my favorite journalists, John Hersey, once noted that:
A writer is bound to have varying degrees of success, and I think that that is partly an issue of how central the burden of the story is to the author’s psyche.
If burdens can translate to the weights of stories, a good example is Hersey’s epochal editorial of Hiroshima, arguably one of the most influential pieces of American writing in the 20th century. After the first atomic bomb was dropped over Hiroshima in August 1945, Hersey set out to uncover the effect of nuclear war not to buildings but to humans. He interviewed several dozen survivors and settled on six regular people who shared powerful memories. His sentences put faces into names. And he used narrative to make those faces alive. Doing so, he unmistakenly bled his journalistic core with a kernel of moral preachment. Hersey’s art of facts did wonder — Albert Einstein, for example, requested one thousand copies of the editorial to distribute to members of his Emergency Committee of Atomic Scientists and noted:
I believe Mr. Hersey has given a true picture of the appalling effect on human beings. . . . And this picture has implications for the future of mankind which must deeply concern all responsible men and women.
Writing with burdens also means writing with a disregard for moral fashion. With the advent of web publishing tools in the 1990s, public writing has been democratized in a dramatic fashion. Everyone write. Just look at twitter. Yet it isn’t hard to notice what people write about is narrowing. Trumps, Kardashians, and Epsteins of the world dominate our screens. Op-ed sections on major newspapers are painfully biased. Different forms of Russel Conjugation in the media exploit the gap between the emotional content of a word and the factual content. Paul Graham wrote about a related thought experiment and called it the Conformist Test:
Do you have any opinions that you would be reluctant to express in front of a group of your peers?
He made a good point, but this test is not entirely representative. Today, we are often expected to speak. Silence is not an option. Or worse, it translates to a statement. When I last went to a family get-together, I was asked how the food was. “They were all delicious,” I responded without thinking. The dishes were great, but that didn’t matter. In cultures like mine, family unity trumps the dishes. This may be trivial, but the point is societal preferences are preconditioned. And in larger spheres of politics and society, this becomes much more dangerous in aggregate.
In 2016, any party in Silicon Valley would have a discussion about how ridiculous a Trump presidency would be. Whether the party has four people, forty, or four hundred, almost everyone would say they are not voting for Trump. Every poll in the country agrees. Why? Because in Silicon Valley and places alike, there are consequences for saying otherwise so the minority changes their preferences to match the greater consensus. This idea is not new. Nor is the claim that the collective misrepresentation can generate outcomes that are very different from the said public opinion. Economist Timur Kuran calls it “Preference Falsification”. In his 1995 book, Private Truths Public Lies, he puts:
Social coercion, both overt and subtle, causes us to “live a lie” — to misrepresent what we really think in an effort to appease others.
Kuran believes widespread preference falsification is particularly dangerous:
It may give rise to multiple soical equilibria. Small events that seem to bespeak a change in public sentiment can gather momentum and lead to dramtical political upheaval en route to a new equilibium, the paradigmatic example being the downfall of communism in Eastern Europe.
An updated Conformist test would be: Do you have any opinions that you would express the opposite of what you actually think in front of a group of your peers?
Recognizing this problem is an important start, but I wonder if the problem itself will get worse. Last month, OpenAI released the API for its third generation of language prediction model (GPT-3) into the open-source wild. From translation, question-answering, to tasks that require on-the-fly reasoning, such as unscrambling words or using a novel word in a sentence, GTP-3 demonstrated AI can solve many tasks that it has never encountered with just enough data. Just on the text-generation front, humans already can’t distinguish between articles written by humans versus articles written by GPT-3.
The danger, however, is that internet-trained models have internet-scaled biases. When AI starts to generate texts, write screenplays, and produce music, we are in real trouble. Skewed, or worse, falsified input will spit out substances that will become an integral part of our internet fabric.
What now? We need to separate the imperative to write about ideas from ideology and politics. Growing up in China and getting my education at a US university like Stanford, I was often asked about their differences. There’s a lot, but to me their similarity is more striking: the same desire for technological progress and the same resistance to change. Writing what you believe isn’t easy, but we need to start somewhere. Novelist Roberto Bolaño once described why writing is simply necessary.
For me, the word “writing” is the exact opposite of the word “waiting”.
So this is my attempt to do just that. With that, I just have one ask: Let’s please make this a conversation. If you disagree, let me know. I really do hope to make it better. And if by some miracle I have clearly communicated the point and we actually disagree, maybe we can stop this writing business and move this over to a 1 on 1 conversation.